All pull requests

Pull requests

• Reset VAD session.update overrides when a session ends

  #640·andimarafioti opened Oct 1, 2026Member·

  ·
  1

• Break the turn latency line into the waits it was hiding

  #635·salignatmoandal opened Sep 30, 2026Contributor·

  ·

• Keep the configured Kokoro voice for replies in the setup language

  #629·breken-ai opened Sep 30, 2026·

  ·
  1

• Resample Pocket TTS output with a stateful filter

  #628·breken-ai opened Sep 30, 2026·

  ·
  1

• Log FasterWhisper timing only for the final transcription

  #626·salignatmoandal opened Sep 29, 2026Contributor·1 task done·

  ·

• Record final STT latency for Qwen3-ASR

  #624·cpoisson opened Sep 29, 2026·

  ·

• feat: add optional Orukeet STT backend

  #623·tavallaie opened Sep 29, 2026Contributor·

  ·
  1

• Use GGML for Qwen3-TTS on macOS and remove its MLX implementation

  #621·andimarafioti opened Sep 29, 2026Member·

  ·

• Open a turn at 192 ms, keep 384 ms for barge-in

  #599·ServeurpersoCom opened Sep 25, 2026Contributor·

  ·
  1

• Add Big Bench Audio system tests on HF Jobs

  #587·Deep-unlearning opened Sep 24, 2026Collaborator·

  ·

• Stop stale token usage from hijacking the live response key

  #586·ousamabenyounes opened Sep 24, 2026Contributor·

  ·
  2

• Add per-language reference voices for OmniVoice cloning

  #584·nepyope opened Sep 23, 2026·

  ·

• Add Nemotron 3 speaker diarization to live conversations

  #583·andimarafioti opened Sep 23, 2026Member·

  ·
  16

• Stop sending idle WebRTC audio packets

  #570·ousamabenyounes opened Sep 19, 2026Contributor·

  ·
  1

  6

• Fix Parakeet language reporting: distinguish pinned from detected

  #567·alexdimmock95 opened Sep 18, 2026Contributor·

  ·
  1

• fix: keep active response playing on listener backchannels

  #566·xprilion opened Sep 16, 2026Contributor·7 tasks done·

  ·
  1

  1

• feat(responses): keep reasoning items across tool continuations

  #564·modelpath-dev opened Sep 16, 2026·

  ·
  1

• Support admitted mid-session STT, LLM, and TTS model updates

  #548·andimarafioti opened Sep 5, 2026Member·

  ·
  1

  1

• Feat/vision resolver #313

  #402·Meet-Shah10 opened Jul 31, 2026·

  ·

• feat(llm): add short spoken filler sentences for delayed LLM responses (#315)

  #383·AyushDubey23 opened Jul 29, 2026·

  ·
  1

• Fix faster-whisper adapt_gen_kwargs consuming its caller's config dict

  #374·LHMQ878 opened Jul 28, 2026·

  ·
  1

• Fix language leaking between Whisper STT handlers via a shared gen_kwargs default

  #373·LHMQ878 opened Jul 28, 2026·

  ·
  1

• Add Telnyx managed STT and TTS backends

  #372·a692570 opened Jul 28, 2026·

  ·

• Add voice cloning for the Qwen3 TTS backend

  #371·A-Mahla opened Jul 28, 2026Contributor·

  ·

• fix: RuntimeError: Cannot execute emptyCache() without MPS backend.

  #370·thiagobarbosa opened Jul 24, 2026·

  ·
  1

  1